Investigating Spire GNSS RO Bending Angle Assimilation Impacts on HWRF Forecasts of Four 2022 Atlantic Hurricanes

Global Navigation Satellite System (GNSS) radio occultation (RO) signals received by low earth orbit (LEO) receiver satellites can be processed to retrieve RO bending angles, providing information about atmospheric water vapor, temperature, and pressure profiles along the RO ray limb soundings. Due to their all-sky observability, high vertical resolution, and deep tropospheric penetration, assimilated GNSS RO observations can improve numerical weather prediction (NWP) model tropical cyclone (TC) forecasts. While traditionally provided by government-funded missions such as the joint US-Taiwan COSMIC-2 constellation, an increasingly large volume of GNSS RO data from commercial satellites has become available in recent years for NWP model data assimilation (DA). Spire Global, Inc. (hereafter Spire) provided the National Oceanic and Atmospheric Administration (NOAA) with approximately 6,000 daily RO profiles during the 2022 Atlantic hurricane season as part of their Delivery Order-4 (DO-4) contract under NOAA’s multi-year Commercial Weather Data Pilot program. This program is designed to evaluate the impacts of assimilated commercial RO data on NWP forecasts.

To investigate the impacts of Spire RO DA on a regional NWP model that already assimilates COSMIC-2 data, we conducted cycled forecast data denial experiments using the Hurricane Weather Research and Forecasting (HWRF) model for four 2022 Atlantic hurricane cases: Earl, Fiona, Ian, and Julia. The experiments included: (i) C2SPIRE, assimilating the DO-4 Spire RO bending angles on top of all other operationally assimilated RO observations, including those from COSMIC-2; (ii) C2, similar to C2SPIRE but excluding the DO-4 Spire RO observations; and (iii) CONTROL, similar to C2 but without the COSMIC-2 RO observations.

Considering all 108 HWRF forecasts initialized from cycled analyses for the four hurricane cases, our findings indicate that C2SPIRE – and, to a lesser extent, C2 – show a statistically significant reduced over-intensification bias in minimum central sea-level pressure (P_MIN) and maximum surface wind (V_MAX), compared to CONTROL, particularly for medium-to-long-range forecast verification times. These intensity forecast differences are consistent with C2SPIRE short-range forecasts having a reduced lower troposphere specific humidity positive bias and RMSD, as compared to C2 short-range forecasts. These findings are verified using dropsondes released by aircraft reconnaissance missions. Contoured frequency by altitude diagram (CFAD) plots, comparing the full distributions of temperature, specific humidity, and wind speed biases in C2SPIRE, C2, and CONTROL short-range forecasts, reveal additional details into the beneficial impacts of Spire RO DA. Neither the Spire nor COSMIC-2 RO observations substantially impact TC track forecast errors, likely due to the HWRF system’s design which limits DA impacts outside of the vortex region.

In conclusion, our results suggest that despite Spire RO observations’ reduced spatial coverage over the tropics compared to COSMIC-2, they can still provide additional value to NWP model TC forecasts by helping to correct biases in model water vapor forecasts. However, it is important to note that our statistical HWRF error analysis is heavily influenced by two hurricanes – Earl and Fiona – that spent significant portions of their lifecycles in the subtropics. In these regions, the spatial coverage density of Spire and COSMIC-2 observations is more comparable. Therefore, more case studies are needed to evaluate the broader value of Spire RO DA in forecasting tropical cyclones, particularly those primarily confined to the tropics.